1 /* SPDX-License-Identifier: GPL-2.0 */
2 /*
3  * fs/f2fs/node.h
4  *
5  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6  *             http://www.samsung.com/
7  */
8 /* start node id of a node block dedicated to the given node id */
9 #define	START_NID(nid) (((nid) / NAT_ENTRY_PER_BLOCK) * NAT_ENTRY_PER_BLOCK)
10 
11 /* node block offset on the NAT area dedicated to the given start node id */
12 #define	NAT_BLOCK_OFFSET(start_nid) ((start_nid) / NAT_ENTRY_PER_BLOCK)
13 
14 /* # of pages to perform synchronous readahead before building free nids */
15 #define FREE_NID_PAGES	8
16 #define MAX_FREE_NIDS	(NAT_ENTRY_PER_BLOCK * FREE_NID_PAGES)
17 
18 /* size of free nid batch when shrinking */
19 #define SHRINK_NID_BATCH_SIZE	8
20 
21 #define DEF_RA_NID_PAGES	0	/* # of nid pages to be readaheaded */
22 
23 /* maximum readahead size for node during getting data blocks */
24 #define MAX_RA_NODE		128
25 
26 /* control the memory footprint threshold (10MB per 1GB ram) */
27 #define DEF_RAM_THRESHOLD	1
28 
29 /* control dirty nats ratio threshold (default: 10% over max nid count) */
30 #define DEF_DIRTY_NAT_RATIO_THRESHOLD		10
31 /* control total # of nats */
32 #define DEF_NAT_CACHE_THRESHOLD			100000
33 
34 /* control total # of node writes used for roll-fowrad recovery */
35 #define DEF_RF_NODE_BLOCKS			0
36 
37 /* vector size for gang look-up from nat cache that consists of radix tree */
38 #define NATVEC_SIZE	64
39 #define SETVEC_SIZE	32
40 
41 /* return value for read_node_page */
42 #define LOCKED_PAGE	1
43 
44 /* check pinned file's alignment status of physical blocks */
45 #define FILE_NOT_ALIGNED	1
46 
47 /* For flag in struct node_info */
48 enum {
49 	IS_CHECKPOINTED,	/* is it checkpointed before? */
50 	HAS_FSYNCED_INODE,	/* is the inode fsynced before? */
51 	HAS_LAST_FSYNC,		/* has the latest node fsync mark? */
52 	IS_DIRTY,		/* this nat entry is dirty? */
53 	IS_PREALLOC,		/* nat entry is preallocated */
54 };
55 
56 /*
57  * For node information
58  */
59 struct node_info {
60 	nid_t nid;		/* node id */
61 	nid_t ino;		/* inode number of the node's owner */
62 	block_t	blk_addr;	/* block address of the node */
63 	unsigned char version;	/* version of the node */
64 	unsigned char flag;	/* for node information bits */
65 };
66 
67 struct nat_entry {
68 	struct list_head list;	/* for clean or dirty nat list */
69 	struct node_info ni;	/* in-memory node information */
70 };
71 
72 #define nat_get_nid(nat)		((nat)->ni.nid)
73 #define nat_set_nid(nat, n)		((nat)->ni.nid = (n))
74 #define nat_get_blkaddr(nat)		((nat)->ni.blk_addr)
75 #define nat_set_blkaddr(nat, b)		((nat)->ni.blk_addr = (b))
76 #define nat_get_ino(nat)		((nat)->ni.ino)
77 #define nat_set_ino(nat, i)		((nat)->ni.ino = (i))
78 #define nat_get_version(nat)		((nat)->ni.version)
79 #define nat_set_version(nat, v)		((nat)->ni.version = (v))
80 
81 #define inc_node_version(version)	(++(version))
82 
copy_node_info(struct node_info * dst,struct node_info * src)83 static inline void copy_node_info(struct node_info *dst,
84 						struct node_info *src)
85 {
86 	dst->nid = src->nid;
87 	dst->ino = src->ino;
88 	dst->blk_addr = src->blk_addr;
89 	dst->version = src->version;
90 	/* should not copy flag here */
91 }
92 
set_nat_flag(struct nat_entry * ne,unsigned int type,bool set)93 static inline void set_nat_flag(struct nat_entry *ne,
94 				unsigned int type, bool set)
95 {
96 	unsigned char mask = 0x01 << type;
97 	if (set)
98 		ne->ni.flag |= mask;
99 	else
100 		ne->ni.flag &= ~mask;
101 }
102 
get_nat_flag(struct nat_entry * ne,unsigned int type)103 static inline bool get_nat_flag(struct nat_entry *ne, unsigned int type)
104 {
105 	unsigned char mask = 0x01 << type;
106 	return ne->ni.flag & mask;
107 }
108 
nat_reset_flag(struct nat_entry * ne)109 static inline void nat_reset_flag(struct nat_entry *ne)
110 {
111 	/* these states can be set only after checkpoint was done */
112 	set_nat_flag(ne, IS_CHECKPOINTED, true);
113 	set_nat_flag(ne, HAS_FSYNCED_INODE, false);
114 	set_nat_flag(ne, HAS_LAST_FSYNC, true);
115 }
116 
node_info_from_raw_nat(struct node_info * ni,struct f2fs_nat_entry * raw_ne)117 static inline void node_info_from_raw_nat(struct node_info *ni,
118 						struct f2fs_nat_entry *raw_ne)
119 {
120 	ni->ino = le32_to_cpu(raw_ne->ino);
121 	ni->blk_addr = le32_to_cpu(raw_ne->block_addr);
122 	ni->version = raw_ne->version;
123 }
124 
raw_nat_from_node_info(struct f2fs_nat_entry * raw_ne,struct node_info * ni)125 static inline void raw_nat_from_node_info(struct f2fs_nat_entry *raw_ne,
126 						struct node_info *ni)
127 {
128 	raw_ne->ino = cpu_to_le32(ni->ino);
129 	raw_ne->block_addr = cpu_to_le32(ni->blk_addr);
130 	raw_ne->version = ni->version;
131 }
132 
excess_dirty_nats(struct f2fs_sb_info * sbi)133 static inline bool excess_dirty_nats(struct f2fs_sb_info *sbi)
134 {
135 	return NM_I(sbi)->nat_cnt[DIRTY_NAT] >= NM_I(sbi)->max_nid *
136 					NM_I(sbi)->dirty_nats_ratio / 100;
137 }
138 
excess_cached_nats(struct f2fs_sb_info * sbi)139 static inline bool excess_cached_nats(struct f2fs_sb_info *sbi)
140 {
141 	return NM_I(sbi)->nat_cnt[TOTAL_NAT] >= DEF_NAT_CACHE_THRESHOLD;
142 }
143 
144 enum mem_type {
145 	FREE_NIDS,	/* indicates the free nid list */
146 	NAT_ENTRIES,	/* indicates the cached nat entry */
147 	DIRTY_DENTS,	/* indicates dirty dentry pages */
148 	INO_ENTRIES,	/* indicates inode entries */
149 	EXTENT_CACHE,	/* indicates extent cache */
150 	DISCARD_CACHE,	/* indicates memory of cached discard cmds */
151 	COMPRESS_PAGE,	/* indicates memory of cached compressed pages */
152 	BASE_CHECK,	/* check kernel status */
153 };
154 
155 struct nat_entry_set {
156 	struct list_head set_list;	/* link with other nat sets */
157 	struct list_head entry_list;	/* link with dirty nat entries */
158 	nid_t set;			/* set number*/
159 	unsigned int entry_cnt;		/* the # of nat entries in set */
160 };
161 
162 struct free_nid {
163 	struct list_head list;	/* for free node id list */
164 	nid_t nid;		/* node id */
165 	int state;		/* in use or not: FREE_NID or PREALLOC_NID */
166 };
167 
next_free_nid(struct f2fs_sb_info * sbi,nid_t * nid)168 static inline void next_free_nid(struct f2fs_sb_info *sbi, nid_t *nid)
169 {
170 	struct f2fs_nm_info *nm_i = NM_I(sbi);
171 	struct free_nid *fnid;
172 
173 	spin_lock(&nm_i->nid_list_lock);
174 	if (nm_i->nid_cnt[FREE_NID] <= 0) {
175 		spin_unlock(&nm_i->nid_list_lock);
176 		return;
177 	}
178 	fnid = list_first_entry(&nm_i->free_nid_list, struct free_nid, list);
179 	*nid = fnid->nid;
180 	spin_unlock(&nm_i->nid_list_lock);
181 }
182 
183 /*
184  * inline functions
185  */
get_nat_bitmap(struct f2fs_sb_info * sbi,void * addr)186 static inline void get_nat_bitmap(struct f2fs_sb_info *sbi, void *addr)
187 {
188 	struct f2fs_nm_info *nm_i = NM_I(sbi);
189 
190 #ifdef CONFIG_F2FS_CHECK_FS
191 	if (memcmp(nm_i->nat_bitmap, nm_i->nat_bitmap_mir,
192 						nm_i->bitmap_size))
193 		f2fs_bug_on(sbi, 1);
194 #endif
195 	memcpy(addr, nm_i->nat_bitmap, nm_i->bitmap_size);
196 }
197 
current_nat_addr(struct f2fs_sb_info * sbi,nid_t start)198 static inline pgoff_t current_nat_addr(struct f2fs_sb_info *sbi, nid_t start)
199 {
200 	struct f2fs_nm_info *nm_i = NM_I(sbi);
201 	pgoff_t block_off;
202 	pgoff_t block_addr;
203 
204 	/*
205 	 * block_off = segment_off * 512 + off_in_segment
206 	 * OLD = (segment_off * 512) * 2 + off_in_segment
207 	 * NEW = 2 * (segment_off * 512 + off_in_segment) - off_in_segment
208 	 */
209 	block_off = NAT_BLOCK_OFFSET(start);
210 
211 	block_addr = (pgoff_t)(nm_i->nat_blkaddr +
212 		(block_off << 1) -
213 		(block_off & (sbi->blocks_per_seg - 1)));
214 
215 	if (f2fs_test_bit(block_off, nm_i->nat_bitmap))
216 		block_addr += sbi->blocks_per_seg;
217 
218 	return block_addr;
219 }
220 
next_nat_addr(struct f2fs_sb_info * sbi,pgoff_t block_addr)221 static inline pgoff_t next_nat_addr(struct f2fs_sb_info *sbi,
222 						pgoff_t block_addr)
223 {
224 	struct f2fs_nm_info *nm_i = NM_I(sbi);
225 
226 	block_addr -= nm_i->nat_blkaddr;
227 	block_addr ^= 1 << sbi->log_blocks_per_seg;
228 	return block_addr + nm_i->nat_blkaddr;
229 }
230 
set_to_next_nat(struct f2fs_nm_info * nm_i,nid_t start_nid)231 static inline void set_to_next_nat(struct f2fs_nm_info *nm_i, nid_t start_nid)
232 {
233 	unsigned int block_off = NAT_BLOCK_OFFSET(start_nid);
234 
235 	f2fs_change_bit(block_off, nm_i->nat_bitmap);
236 #ifdef CONFIG_F2FS_CHECK_FS
237 	f2fs_change_bit(block_off, nm_i->nat_bitmap_mir);
238 #endif
239 }
240 
ino_of_node(struct page * node_page)241 static inline nid_t ino_of_node(struct page *node_page)
242 {
243 	struct f2fs_node *rn = F2FS_NODE(node_page);
244 	return le32_to_cpu(rn->footer.ino);
245 }
246 
nid_of_node(struct page * node_page)247 static inline nid_t nid_of_node(struct page *node_page)
248 {
249 	struct f2fs_node *rn = F2FS_NODE(node_page);
250 	return le32_to_cpu(rn->footer.nid);
251 }
252 
ofs_of_node(struct page * node_page)253 static inline unsigned int ofs_of_node(struct page *node_page)
254 {
255 	struct f2fs_node *rn = F2FS_NODE(node_page);
256 	unsigned flag = le32_to_cpu(rn->footer.flag);
257 	return flag >> OFFSET_BIT_SHIFT;
258 }
259 
cpver_of_node(struct page * node_page)260 static inline __u64 cpver_of_node(struct page *node_page)
261 {
262 	struct f2fs_node *rn = F2FS_NODE(node_page);
263 	return le64_to_cpu(rn->footer.cp_ver);
264 }
265 
next_blkaddr_of_node(struct page * node_page)266 static inline block_t next_blkaddr_of_node(struct page *node_page)
267 {
268 	struct f2fs_node *rn = F2FS_NODE(node_page);
269 	return le32_to_cpu(rn->footer.next_blkaddr);
270 }
271 
fill_node_footer(struct page * page,nid_t nid,nid_t ino,unsigned int ofs,bool reset)272 static inline void fill_node_footer(struct page *page, nid_t nid,
273 				nid_t ino, unsigned int ofs, bool reset)
274 {
275 	struct f2fs_node *rn = F2FS_NODE(page);
276 	unsigned int old_flag = 0;
277 
278 	if (reset)
279 		memset(rn, 0, sizeof(*rn));
280 	else
281 		old_flag = le32_to_cpu(rn->footer.flag);
282 
283 	rn->footer.nid = cpu_to_le32(nid);
284 	rn->footer.ino = cpu_to_le32(ino);
285 
286 	/* should remain old flag bits such as COLD_BIT_SHIFT */
287 	rn->footer.flag = cpu_to_le32((ofs << OFFSET_BIT_SHIFT) |
288 					(old_flag & OFFSET_BIT_MASK));
289 }
290 
copy_node_footer(struct page * dst,struct page * src)291 static inline void copy_node_footer(struct page *dst, struct page *src)
292 {
293 	struct f2fs_node *src_rn = F2FS_NODE(src);
294 	struct f2fs_node *dst_rn = F2FS_NODE(dst);
295 	memcpy(&dst_rn->footer, &src_rn->footer, sizeof(struct node_footer));
296 }
297 
fill_node_footer_blkaddr(struct page * page,block_t blkaddr)298 static inline void fill_node_footer_blkaddr(struct page *page, block_t blkaddr)
299 {
300 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_P_SB(page));
301 	struct f2fs_node *rn = F2FS_NODE(page);
302 	__u64 cp_ver = cur_cp_version(ckpt);
303 
304 	if (__is_set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG))
305 		cp_ver |= (cur_cp_crc(ckpt) << 32);
306 
307 	rn->footer.cp_ver = cpu_to_le64(cp_ver);
308 	rn->footer.next_blkaddr = cpu_to_le32(blkaddr);
309 }
310 
is_recoverable_dnode(struct page * page)311 static inline bool is_recoverable_dnode(struct page *page)
312 {
313 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_P_SB(page));
314 	__u64 cp_ver = cur_cp_version(ckpt);
315 
316 	/* Don't care crc part, if fsck.f2fs sets it. */
317 	if (__is_set_ckpt_flags(ckpt, CP_NOCRC_RECOVERY_FLAG))
318 		return (cp_ver << 32) == (cpver_of_node(page) << 32);
319 
320 	if (__is_set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG))
321 		cp_ver |= (cur_cp_crc(ckpt) << 32);
322 
323 	return cp_ver == cpver_of_node(page);
324 }
325 
326 /*
327  * f2fs assigns the following node offsets described as (num).
328  * N = NIDS_PER_BLOCK
329  *
330  *  Inode block (0)
331  *    |- direct node (1)
332  *    |- direct node (2)
333  *    |- indirect node (3)
334  *    |            `- direct node (4 => 4 + N - 1)
335  *    |- indirect node (4 + N)
336  *    |            `- direct node (5 + N => 5 + 2N - 1)
337  *    `- double indirect node (5 + 2N)
338  *                 `- indirect node (6 + 2N)
339  *                       `- direct node
340  *                 ......
341  *                 `- indirect node ((6 + 2N) + x(N + 1))
342  *                       `- direct node
343  *                 ......
344  *                 `- indirect node ((6 + 2N) + (N - 1)(N + 1))
345  *                       `- direct node
346  */
IS_DNODE(struct page * node_page)347 static inline bool IS_DNODE(struct page *node_page)
348 {
349 	unsigned int ofs = ofs_of_node(node_page);
350 
351 	if (f2fs_has_xattr_block(ofs))
352 		return true;
353 
354 	if (ofs == 3 || ofs == 4 + NIDS_PER_BLOCK ||
355 			ofs == 5 + 2 * NIDS_PER_BLOCK)
356 		return false;
357 	if (ofs >= 6 + 2 * NIDS_PER_BLOCK) {
358 		ofs -= 6 + 2 * NIDS_PER_BLOCK;
359 		if (!((long int)ofs % (NIDS_PER_BLOCK + 1)))
360 			return false;
361 	}
362 	return true;
363 }
364 
set_nid(struct page * p,int off,nid_t nid,bool i)365 static inline int set_nid(struct page *p, int off, nid_t nid, bool i)
366 {
367 	struct f2fs_node *rn = F2FS_NODE(p);
368 
369 	f2fs_wait_on_page_writeback(p, NODE, true, true);
370 
371 	if (i)
372 		rn->i.i_nid[off - NODE_DIR1_BLOCK] = cpu_to_le32(nid);
373 	else
374 		rn->in.nid[off] = cpu_to_le32(nid);
375 	return set_page_dirty(p);
376 }
377 
get_nid(struct page * p,int off,bool i)378 static inline nid_t get_nid(struct page *p, int off, bool i)
379 {
380 	struct f2fs_node *rn = F2FS_NODE(p);
381 
382 	if (i)
383 		return le32_to_cpu(rn->i.i_nid[off - NODE_DIR1_BLOCK]);
384 	return le32_to_cpu(rn->in.nid[off]);
385 }
386 
387 /*
388  * Coldness identification:
389  *  - Mark cold files in f2fs_inode_info
390  *  - Mark cold node blocks in their node footer
391  *  - Mark cold data pages in page cache
392  */
393 
is_node(struct page * page,int type)394 static inline int is_node(struct page *page, int type)
395 {
396 	struct f2fs_node *rn = F2FS_NODE(page);
397 	return le32_to_cpu(rn->footer.flag) & (1 << type);
398 }
399 
400 #define is_cold_node(page)	is_node(page, COLD_BIT_SHIFT)
401 #define is_fsync_dnode(page)	is_node(page, FSYNC_BIT_SHIFT)
402 #define is_dent_dnode(page)	is_node(page, DENT_BIT_SHIFT)
403 
set_cold_node(struct page * page,bool is_dir)404 static inline void set_cold_node(struct page *page, bool is_dir)
405 {
406 	struct f2fs_node *rn = F2FS_NODE(page);
407 	unsigned int flag = le32_to_cpu(rn->footer.flag);
408 
409 	if (is_dir)
410 		flag &= ~(0x1 << COLD_BIT_SHIFT);
411 	else
412 		flag |= (0x1 << COLD_BIT_SHIFT);
413 	rn->footer.flag = cpu_to_le32(flag);
414 }
415 
set_mark(struct page * page,int mark,int type)416 static inline void set_mark(struct page *page, int mark, int type)
417 {
418 	struct f2fs_node *rn = F2FS_NODE(page);
419 	unsigned int flag = le32_to_cpu(rn->footer.flag);
420 	if (mark)
421 		flag |= (0x1 << type);
422 	else
423 		flag &= ~(0x1 << type);
424 	rn->footer.flag = cpu_to_le32(flag);
425 
426 #ifdef CONFIG_F2FS_CHECK_FS
427 	f2fs_inode_chksum_set(F2FS_P_SB(page), page);
428 #endif
429 }
430 #define set_dentry_mark(page, mark)	set_mark(page, mark, DENT_BIT_SHIFT)
431 #define set_fsync_mark(page, mark)	set_mark(page, mark, FSYNC_BIT_SHIFT)
432